Is it right to conclude that pdns doesn't require a serial
number in the SOA field if working in a sql replication setup?
I added my zones with only 'nsa.byte.nl' as SOA content. The nameserver
check of the Dutch domain registry  seems to accept this, except for
some weird cases. If pdns lacks a serial number in the SOA record, it
seems to make one itself. But it looks like it does that at random (?)
intervals. The serials are increased every minute or so. In my situation
this gives a problem because sometimes the serial of the 2nd nameserver is
higher than the 1st.
Is it fundamentally right to leave out the serial numbers in the SOA
If so, how do I sync the updating of serial numbers between the two
nameservers (that use replicated mysql databases).
On Mon, Mar 03, 2003 at 05:50:13PM +0100, Willem de Groot wrote:
> Is it fundamentally right to leave out the serial numbers in the SOA
no, please don't do that right now. There is some 'magic' code around that
could help you automate serial number generation. But for now, please fill
out a serial number. You can leave out the rest of the SOA though.
> If so, how do I sync the updating of serial numbers between the two
> nameservers (that use replicated mysql databases).
Just update the master, the slave will follow automatically over MySQL
Hi, sorry for spamming this list again, it's just that I'm really excited
about this superslave/master feature which I can't get to work.
I set up a superslave/master config, with:
- pdns 2.9.6 as superslave
- bind 9.1.2 as supermaster
But the following errors occur on the slave:
pdns: [ID 702911 daemon.notice] No serial for 'testbyte.nl' found -
zone is missing?
pdns: [ID 702911 daemon.error] Unable to AXFR zone 'testbyte.nl':
Trying to read data from remote TCP client: Connection reset by peer
This first line looks correct -- it just received a notify from the master
and hasn't got a serial yet. But while axfr'ing the zone, the transfer is
I double checked my firewall setup and I'm really sure that that is not
the problem. My packet log on the slave says:
p SL,40415 -> MA,53 PR tcp len 20 48 -S OUT
p MA,53 -> SL,40415 PR tcp len 20 48 -AS IN
p SL,40415 -> MA,53 PR tcp len 20 40 -A OUT
p SL,40415 -> MA,53 PR tcp len 20 71 -AP OUT
p MA,53 -> SL,40415 PR tcp len 20 40 -A IN
p MA,53 -> SL,40415 PR tcp len 20 40 -AR IN
Where SL is the slave-ip, MA the master-ip.
According to my firewall log, pdns sends out TCP connections from another
ip than set with 'local-address' in pdns.conf. I don't think this is the
cause for my problem, because the master is set up to allow zone transfers
to both ip's. But maybe it influences it in another way?
I can't find anything suspicious in my master-log.
I'm really out of clues. Maybe you have a suggestion?
> According to my firewall log, pdns sends out TCP connections from another
> ip than set with 'local-address' in pdns.conf. I don't think this is the
> cause for my problem, because the master is set up to allow zone transfers
> to both ip's. But maybe it influences it in another way?
I 'fixed' that last week with a patch that adds a 'query-local-address'
option that allows you to specify a source IP address for outgoing
queries (including AXFRs).
You can wait for the next release or just apply the patch from